Background. A self-rated measure of health anxiety should be sensitive across the full range of intensity (from mild concern to frank hypochondriasis) and should differentiate people suffering from health anxiety from those who have actual physical illness but who are not excessively concerned about their health. It should also encompass the full range of clinical symptoms characteristic of clinical hypochondriasis. The development and validation of such a scale is described.

Method. Three studies were conducted. First, the questionnaire was validated by comparing the responses of patients suffering from hypochondriasis with those suffering from hypochondriasis and panic disorder, panic disorder, social phobia and non-patient controls. Secondly, a state version of the questionnaire was administered to patients undergoing cognitive-behavioural treatment or wait-list in order to examine the measure's sensitivity to change. In the third study, a shortened version was developed and validated in similar types of sample, and in a range of samples of people seeking medical help for physical illness.

Results. The scale was found to be reliable and to have a high internal consistency. Hypochondriacal patients scored significantly higher than anxiety disorder patients, including both social phobic patients and panic disorder patients as well as normal controls. In the second study, a ‘state’ version of the scale was found to be sensitive to treatment effects, and to correlate very highly with a clinician rating based on an interview of present clinical state. A development and refinement of the scale (intended to reflect more fully the range of symptoms of and reactions to hypochondriasis) was found to be reliable and valid. A very short (14 item) version of the scale was found to have comparable properties to the full length scale.

Conclusions. The HAI is a reliable and valid measure of health anxiety. It is likely to be useful as a brief screening instrument, as there is a short form which correlates highly with the longer version.